[mcp-analysis] MCP Structural Analysis - 2026-04-24

[github-actions[bot]]

9 GitHub MCP tools analyzed across 8 toolsets. Average usefulness rating: 3.0/5. Best tool: list_discussions (5/5 — compact, paginated, complete). Worst: get_me (blocked by integration scope) and list_issues (filtered by integrity policy). Key finding: search_code responses include 30+ URL templates per repository object, list_code_scanning_alerts embeds full markdown help text (~1500 chars/alert), and list_workflows ignores per_page=1 — these three toolsets represent the highest context-waste risk for agents.

Full Structural Analysis Report

Executive Summary

Metric	Value
Tools Analyzed	9
Total Tokens (Today)	12,630
Average Usefulness Rating	3.0/5
Best Rated Tool	list_discussions — 5/5
Worst Rated Tool	get_me / list_issues — 1/5 (env constraints)
Analysis Date	2026-04-24
Run	§24888785593

---

Usefulness Ratings for Agentic Work

Tool	Toolset	Rating	Assessment
list_discussions	discussions	⭐⭐⭐⭐⭐	Excellent: compact, paginated, totalCount included
list_pull_requests	pull_requests	⭐⭐⭐⭐	Good: rich PR data, nested base/head repo details
get_file_contents	repos	⭐⭐⭐⭐	Good: appropriate verbosity; size depends on file
list_label	labels	⭐⭐⭐⭐	Good: clean flat structure; fetches all labels at once
list_workflows	actions	⭐⭐⭐	Adequate: per_page param ignored — returns ~29 items despite requesting 1
list_code_scanning_alerts	code_security	⭐⭐⭐	Adequate: embeds full markdown help text (~1500 chars/alert) — major context bloat
search_code	search	⭐⭐	Limited: repository sub-object has 30+ URL templates agents never use; no text_matches
get_me	context	⭐	Poor: 403 Forbidden — integration token lacks user scope; fails entirely
list_issues	issues	⭐	Poor (env): blocked by integrity policy in this sandbox; 0 items visible

---

Schema Analysis

Tool	Type	Depth	Key Fields	Notable
list_discussions	object	2	number, title, category, html_url, pageInfo, totalCount	Best structure overall
list_pull_requests	array	3	number, title, state, body, labels, user, base, head	Body included (can be large)
get_file_contents	object	2	mimeType, text, uri, sha	SHA verification included
list_label	object	1	id, name, color, description, totalCount	Flat and clean
list_workflows	object	2	id, name, path, state, html_url, total_count	Clean but pagination broken
list_code_scanning_alerts	array	4	number, rule.{id,description,help}, state, most_recent_instance	Help field is ~1500 chars each
search_code	object	4	total_count, items[].{name, path, sha, html_url, repository{...}}	30+ url templates in repo sub-object
get_me	error	1	error_message	Cannot retrieve user profile
list_issues	array	0	(none visible)	Integrity policy blocks all results

---

Response Size Analysis

Toolset	Tokens	Notes
actions	3,500	Per-page parameter ignored
repos	3,000	Scales with file size
code_security	2,500	Help text bloat
labels	2,000	Fetches all labels
search	1,000	URL template bloat
pull_requests	450	Reasonable
discussions	150	Best in class
context	20	Error response
issues	10	Blocked
Total	12,630

---

Recommendations

High-value tools (rating 4-5) — prefer these:

• list_discussions — most context-efficient with complete data
• list_pull_requests — rich data, good for PR workflows
• get_file_contents — appropriate for targeted file reads
• list_label — clean, use for label maps/lookups

Tools needing improvement:

• search_code — strip URL templates from repository sub-object; add text_matches support
• list_code_scanning_alerts — make rule.help opt-in (it's rarely needed inline)
• list_workflows — fix per_page pagination parameter handling
• get_me — document the required token scope clearly

Context-efficient choices (low tokens, high rating):

• list_discussions (150 tokens, rating 5) ← optimal
• list_pull_requests (450 tokens, rating 4) ← good

Context-heavy — use sparingly:

• list_workflows (3,500 tokens) — pagination broken, over-fetches
• get_file_contents (3,000 tokens) — inherently large; choose files carefully
• list_code_scanning_alerts (2,500 tokens) — help text inflates 5× vs data alone

---

Visualizations

Response Size by Toolset

Usefulness Ratings

Token Size vs Usefulness

Nesting Depth by Toolset

References: §24888785593

Note

🔒 Integrity filter blocked 1 item

The following item was blocked because it doesn't meet the GitHub integrity level.

• shared/apm.md restore step crashes on PRs that modify tracked files (e.g. apm.lock.yaml) #28256 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

Generated by GitHub MCP Structural Analysis · ● 370.6K · ◷

• expires on Apr 25, 2026, 12:28 PM UTC